Automatic Speech Recognition
Also written: ASR
Technology that converts spoken language into text.
Automatic Speech Recognition, or ASR, is technology that turns spoken words into written text. Think of it like a digital scribe that listens to you and types out what you say.
This is the same technology that powers voice assistants like Siri or Alexa, allowing them to understand your commands. It's also used for transcribing meetings or videos, making content more accessible and searchable.
Automatic Speech Recognition (ASR) is the process of converting spoken audio into a machine-readable text format. Modern ASR systems typically employ deep neural networks, often based on recurrent neural networks (RNNs) or transformer architectures, trained on vast datasets of transcribed speech. These models learn to map acoustic features extracted from audio signals to phonetic units and then to words.
Key challenges include handling variations in accent, background noise, speaking rate, and language. Tradeoffs exist between model complexity, computational cost, and accuracy across different languages and acoustic conditions. State-of-the-art systems can achieve very high word error rates on clean speech in well-resourced languages.
Meta AI recently announced an "Omnilingual" Automatic Speech Recognition system capable of transcribing over 1,600 languages. This development highlights the ongoing push in frontier AI to expand language support beyond common languages, making speech-to-text technology more accessible globally.
- Introducing Meta Omnilingual Automatic Speech Recognition | Transcription for 1,600+ languagesMeta · AI at Meta on YouTube ·
- Meta Meta AI
- Meta AI has developed an "Omnilingual" Automatic Speech Recognition system that supports over 1,600 languages, aiming to cover many previously unserved languages.
- Speech Synthesis
- This is the opposite of ASR; it converts text into spoken word
- Natural Language Understanding
- This focuses on interpreting the meaning of text
- Speaker Recognition
- This identifies who is speaking, rather than what is being
The concept of automatic speech recognition has roots in early artificial intelligence and signal processing research dating back to the mid-20th century. Early systems were rule-based, but modern approaches are dominated by machine learning.
Open the voice typing feature on your smartphone or computer. Speak a few sentences and watch them appear as text. Then, try speaking in a different accent or with some background noise to see how the system handles variations.
The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.
- PromptRole-play a student struggling with math for AI tutor simulationSimulate a student's math struggles and uncertainty for AI tutor evaluation. · 5 min
- TemplateReusable workflow templates for healthcare teamsTeams can create and share templates for common workflows like documentation or prior authorizations. · 15 min
- TemplateTemplate for building an AI tool tailored to your workA structured approach to defining and building a custom AI tool for specific business needs. · 15 min
- TemplateEvals CSV prompt list for agent skillsA CSV file template for listing prompts used in evaluating agent skills. · 20 min
- HeuristicDouble coding agent task completion with docs linksProviding coding agents with a documentation link significantly increases their task completion rate. · 2 min
Questions people ask
- How accurate is ASR?
- Accuracy varies greatly depending on the system, language, audio quality, and accents. State-of-the-art systems can be very accurate for common languages in quiet environments.
- Can ASR handle multiple languages at once?
- Some advanced ASR systems are designed to be multilingual and can detect and transcribe different languages within the same audio stream.
First seen on Subvolts Nov 10, 2025. Updated Oct 19, 2026. Written by an AI model from the items above and reviewed against them weekly; report an error.